Presently, network acceleration efforts are blind to user interactions at the UI level. In order to optimize and accelerate network connections, the acceleration systems will typically use predictive algorithms and techniques to determine what a user is doing or is going to do at the UI level. Accordingly, such predictions are inaccurate and flawed, and as such have limited benefit. Thus, improvements in the art are needed.